Llama 4: Meta Platforms (META) is anticipated to launch its new large language model ‘Llama 4’ later this month

Following at least two previous delays, according to a report from The Information on Friday, the company is striving to take the lead in the AI sector.

Nonetheless, the release of Llama 4 might be postponed once more, the report indicated, citing sources who are familiar with the situation.

Major tech companies have been heavily investing in AI infrastructure since the emergence of OpenAI’s ChatGPT, which transformed the technology landscape and spurred investment in machine learning.

The report mentioned that one reason for the delays is that Llama 4 fell short of Meta’s expectations in certain technical benchmarks, especially regarding reasoning and mathematical tasks.

It was also noted that the firm was worried that Llama 4 might not perform as well as OpenAI’s models in facilitating human-like voice conversations.

Under increasing pressure from investors to demonstrate returns on investments, Meta is planning to allocate up to $65 billion this year to enhance its AI infrastructure.

Additionally, the emergence of a popular, cost-effective model from a Chinese tech company, DeepSeek, calls into question the notion that creating the top AI model necessitates billions in funding.

Expertise of Meta Llama 4

According to the report, Llama 4 is expected to incorporate some technical elements from DeepSeek, with at least one version anticipated to utilize a machine-learning technique known as the mixture of experts method, which trains different parts of models for specialized tasks, enabling them to excel in those areas.

Meta is also considering initially launching Llama 4 through Meta AI and subsequently releasing it as open-source software, the report stated.

Last year, Meta introduced the largely free Llama 3 AI model, which can engage in conversations in eight languages, produce higher-quality computer code, and tackle more complex mathematical problems compared to earlier versions.